- Email: [email protected]
Analysis of three years of boundary layer observations over the Gulf of Mexico and its shores
Estuarine, Coastal and Shelf Science 70 (2006) 541e550 www.elsevier.com/locate/ecss
Analysis of three years of boundary layer observations over the Gulf of Mexico and its shores Steven R. Hanna a,*, Clinton P. MacDonald b, Mark Lilly b,1, Charles Knoderer b, C.H. Huang c a Hanna Consultants, 7 Crescent Avenue, Kennebunkport, ME 04046-7235, USA Sonoma Technology, Inc., 1360 Redwood Way, Suite C, Petaluma, CA 94954-1169, USA c Minerals Management Service, 1201 Elmwood Park Blvd., New Orleans, LA 70123, USA b
Received 1 June 2006; accepted 1 June 2006 Available online 6 September 2006
Abstract Boundary layer observations were made over the Gulf of Mexico over a 3-year period in order to develop and test methods for estimating surface fluxes and boundary layer wind fields. In addition to routinely available buoy and CMAN surface data, six 915 MHz radar wind profilers (RWPs) and RASS profilers were mounted on oil platforms and on the shore. Estimates of surface momentum, sensible heat, and latent heat fluxes have been made from the surface observations using the COARE software. Simulations by the National Weather Service’s Eta meteorological model are compared with the observations of surface fluxes and wind profiles. The boundary layer is found to be unstable over 90% of the time, and latent heat fluxes are about five to ten times larger than sensible heat fluxes, as usually found over tropical oceans. Eta model simulations of surface fluxes are within about 50% of COARE estimates of the fluxes based on surface observations. Most of the time, COAREderived fluxes at 11 sites are within a factor of two of each other at any given hour. In multi-day case studies, COARE calculations are found to agree with Eta model simulations of these fluxes and parameters within a factor of two most of the time. Eta model simulations of wind speeds in the boundary layer tend to exceed the RWP observations by 1e2 m s1 near shore and by 2e6 m s1 at distances of 100e200 km offshore. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: marine boundary layer; Gulf of Mexico meteorology; COARE; surface fluxes
1. Introduction This study was driven by the need to calculate the transport and dispersion of pollutants emitted from offshore oil and gas exploration and production activities in the Gulf of Mexico. Much uncertainty exists concerning the atmospheric boundary layer in the region, due to a lack of detailed observations. Complications arise due to space and time variations caused by variable underlying water temperatures and the effects of mesoscale atmospheric phenomena. The current study has involved instrumentation of six offshore oil/gas production platforms to obtain boundary layer observations, including * Corresponding author. E-mail address: [email protected] (S.R. Hanna). 1 Present address: Northrup Grumman Information Technology, 6940 Kings Highway, Alexandria, VA 22319, USA. 0272-7714/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.ecss.2006.06.005
915-MHz Radar Wind Profilers (RWPs), 2-kHz Radio Acoustic Sounding Systems (RASS), and near-surface routine meteorology instruments. Fig. 1 shows a map of the Gulf of Mexico region and indicates the locations of the various observing sites. Two of the profiler sites have operated from May 1998 through September 2001 and four of the sites, sponsored by the Breton Aerometric Monitoring Program (BAMP), operated from October 2000 through September 2001. The RWP measures winds and the RASS measures virtual temperatures between heights of about 100 m and a few kilometers. The near-surface observations at the oil platforms include sea surface temperature as well as wind speed, wind direction, air temperature, and mixing ratio at a reference elevation, zr, of about 25 m (i.e., the platform elevation). These new data, in addition to the traditional data collected by buoys and by the Coastal Marine Automated Network (CMAN), which are available from the National Climatic Data Center (NCDC),
542
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
Fig. 1. Map of the study region in the western and central Gulf of Mexico showing locations of CMAN sites, buoys, and profilers. The two ‘‘MMS’’ profilers operated for about 3 years, while the four BAMP profilers operated for about 1 year. The numbers in parentheses are the water depth in meters.
have been analyzed to investigate the over-water surface energy balance and boundary layer structure for both steady-state horizontally homogeneous conditions and for conditions variable in time and space. These three-dimensional, timedependent fields are being used for analysis of transport and dispersion from over-water sources. The Tropical Ocean Global Atmosphere-Coupled Oceanic Atmospheric Response Experiment (TOGA-COARE) marine boundary layer algorithms (Fairall et al., 1996, 2003) have been used to calculate hourly-averaged and monthly-averaged fundamental boundary layer scaling parameters such as the surface roughness length (zo), the friction velocity (u*), and the MonineObukhov length (L), in addition to the latent heat flux (Qh) and the sensible heat flux (Qs). From these parameters, the mixing depth (h), and the vertical profiles of wind speed (u), temperature (T ), and water vapor mixing ratio (q) can be estimated for input to transport and dispersion models. The basic structure of the COARE marine boundary-layer model is an outgrowth of the LiueKatsaroseBusinger (LKB) (Liu et al., 1979) method. Version 3.0 of COARE, used in the current analysis and described by Fairall et al. (2003), incorporates wave height and period data, in order to increase the accuracy of the estimates of surface fluxes and scaling parameters over shallow areas (Taylor and Yelland, 2000). This allows COARE to account for differences in the characteristics of waves from the deep ocean to the shallow coastal waters. COARE’s accuracy was shown in the model evaluation exercise carried out by Brunke et al. (2003), who evaluated 12 bulk aerodynamic algorithms for computing ocean surface turbulent fluxes with observations from 12 ship cruises over the tropical and midlatitude oceans. They found that COARE
was one of the group of four algorithms that showed the best performance. In the sections below, the outputs of the COARE program (e.g., boundary layer scaling parameters and energy fluxes) are compared to observations and to simulations by the National Center for Environmental Prediction (NCEP) Eta and Eta Data Assimilation System (EDAS) numerical weather prediction modeling system (Black, 1994). In addition, some comparisons are presented of the observed RWP winds and the Eta/EDAS modeled winds. This study focuses on the Western and Central Gulf of Mexico, west of the FloridaeAlabama border. Additional MMS studies (not discussed here) are focusing on the eastern Gulf of Mexico region. 2. Application of the COARE model The Coupled Oceanic Atmospheric Response Experiment (COARE) model (Fairall et al., 1996, 2003) allows estimation of over water boundary layer parameters such as surface fluxes of momentum (t), sensible heat (Qs), and latent heat (Qh), and requires the following input data: Site location and time Wind speed (u), air temperature (Ta), and relative humidity (RH) within the atmospheric surface layer at a reference height zr Sea surface skin temperature (Ts) (or water temperature near the sea surface plus an estimate of solar and net longwave radiation) Mixing height (h)
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
If the water temperature at some level near (i.e., a short distance below) the sea surface is used, then solar and downwelling longwave radiation fluxes need to be estimated from some alternate source in order to correct this temperature to a skin (sea surface) temperature. Precipitation data are not required, but if available, can be used by COARE to estimate the precipitation contribution to the kinetic and thermal energy balance equations. Wave height and period data are not required, but, when available, can be used with Version 3.0 of COARE (Fairall et al., 2003) to account for the different wave structures and theoretically improve the accuracy of the estimates of surface fluxes and scaling parameters over shallow ocean areas. The COARE equations and their justifications are not discussed here, since the focus of the current study is the application of COARE to the extensive data sets collected in the Gulf of Mexico. Fairall et al. (1996, 2003) discuss the details of the COARE equations, and Brunke et al. (2003) show that COARE is one of the better-performing group of models when compared with observations from 12 ship cruises in tropical and midlatitude oceans. Data were acquired and processed from the following sites: 1998 through 2001 offshore buoy data from seven sites, shoreline CMAN station data from five sites, and data collected as part of this project on the Vermillion (VRM), South Marsh Island (SMI), Breton Island (BI), Fort Morgan (FM), Deep Water Platform (DWP), and Shallow Water Platform (SWP) oil/gas production platforms. Fig. 1 shows the locations of the offshore buoys, the CMAN stations, the oil/gas production platforms and the Fort Morgan site (with surface observations plus profilers). The CMAN stations are located on or adjacent to very shallow water on the coast. Since the COARE model is not currently designed to use data collected in such areas, the data from the CMAN sites were used only in test exercises to evaluate how COARE would respond compared to open water sites. The data collected at the platforms generally meet the COARE input data requirements, although the elevations (20e25 m) are relatively high compared to most of the elevations used in development of the model. The data collected at buoys meet most of the COARE data requirements, with the exception that solar and longwave radiation were not observed. Instead, for the buoys, radiation fluxes were estimated using 6-hourly Eta model cloud forecasts and sun elevation data, from which COARE calculates the water skin temperatures from water temperatures at depths of about 0.5e1.0 m. The climatological part of the study is discussed first. Using the hourly meteorological data collected for over 3 years at 11 sites where the data passed QA procedures, COARE was used to calculate hourly sensible heat flux (Qs), latent heat flux (Qh), surface stress (t), friction velocity (u*), temperature scaling parameter (T*), relative humidity scaling parameter (q*), zr/L, and roughness length (zo). Note that zr is the reference height for the measurement and L is the MonineObukhov length. Monthly statistics were then calculated from these hourly values. Monthly averages were not computed if more than 80% of the data in a given month were missing.
543
The case study part of the analysis is discussed second. In this analysis, time series of hourly averages of selected meteorological and derived boundary layer parameters were calculated using COARE, for two multi-day periods during different seasons. The analysis includes comparisons of derived outputs from COARE with simulations of the Eta model for sensible and latent heat flux, and friction velocity. The final part of the analysis involves comparisons of the winds observed by the RWPs at elevations of a few hundred meters with simulations of the Eta model. Mean biases and mean-square errors are presented for the total 3-year period. 3. Results of climatological study with COARE The key results of the Gulf of Mexico boundary layer climatological study (comparisons of monthly averages at 11 sites) with COARE are briefly summarized in this section. The fluxes and scalar parameters calculated by the COARE algorithm for the 3-year observing period in the Gulf of Mexico are physically consistent with our intuitive expectations, and are similar to observations and COARE calculations for the TOGA field study, which took place in the warm Western Pacific Ocean near the equator (Fairall et al., 1996). As expected, the COARE-calculated fluxes in the Gulf of Mexico tend to be strongly related to the sea-air temperature difference. Fig. 2 contains monthly-averaged observed sea-air temperature difference over the 3-year period at seven buoys, two CMAN sites, and the VRM and SMI oil platforms (see Fig. 1 for the locations). An annual variation is seen with maximum differences (about 2e5 C) in December, and minimum differences (about 0.5 C) in May and June. Our analysis shows that the sea-air temperature difference is positive over 90% of the time, implying that well-mixed conditions occur most of the time. The COARE-calculated monthly-average sensible heat fluxes, Qs, over the 3-year period are plotted in Fig. 3 for the same 11 sites used in Fig. 2. Calculated annual-averaged sensible heat fluxes in the Gulf of Mexico average about 10e30 W m2, typical of other over-water sites (Fairall et al., 1996, 2003). Slightly higher sensible heat fluxes (averaging about 30e40 W m2) are calculated in the winter months, especially in December and January, due to larger sea-air temperature differences and stronger winds during the winter. The sensible heat fluxes at the 11 sites tend to agree within a factor of two most of the time during the period from September through March. Larger differences between the 11 sites are seen in Fig. 3 during the summer months, when there are smaller sea-air temperature differences and hence the sensible heat fluxes are relatively small (about 10 W m2 or less) and variable, and are sometimes very near zero or are negative. The COARE-calculated monthly average latent heat fluxes, Qh, are shown in Fig. 4 for the same 11 sites used in Fig. 3. The monthly-averaged latent heat fluxes in Fig. 4 range from about 30 to 260 W m2. The annual average latent heat flux over all 11 sites averages about 50e150 W m2, again typical of other over-water sites. Calculated latent heat fluxes are therefore, on average, about five times larger than the
544
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
Fig. 2. Monthly averaged observed skin (water surface) temperature minus air temperature, for 11 sites in the Gulf of Mexico (see Fig. 1 for locations) for May 1998 through October 2001. The skin temperature was observed only at the platforms (SMI and VRM) and was estimated by COARE from the observed water temperature (at a depth just below the surface) at the other sites.
calculated sensible heat fluxes. This approximate ratio is found at other open-water sites at this latitude (e.g., Fairall et al., 1996). The magnitudes of the fluxes are generally greater at all 11 sites in the late fall months, likely due to the higher water temperatures and the larger sea-air temperature differences and wind speeds at that time. As is well known, the annual
maximum and minimum water temperatures have a time lag of a few months after the maximum and minimum air temperatures. Latent heat fluxes over the Gulf of Mexico are seldom negative due to the fact that the air at the surface is obviously always saturated with water vapor and the sea-air temperature difference is positive 90% of the time.
Fig. 3. Monthly averages of COARE-estimated sensible heat fluxes, for 11 sites in the Gulf of Mexico (see Fig. 1 for locations) for May 1998 through October 2001.
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
545
Fig. 4. Monthly averages of COARE-estimated latent heat fluxes, for 11 sites in the Gulf of Mexico (see Fig. 1 for locations) for May 1998 through October 2001.
Plots of the COARE-calculated monthly average friction velocity, u*, temperature scaling parameter, T*, and humidity scaling parameter, q*, were made but are not given in the current paper due to space limitations. The plots of u* show an agreement among the sites well within a factor of two and often within 20%. This agreement is important because u* is the key scaling velocity for estimating transport speeds and dispersion rates. u* is calculated to be slightly lower from May through August when wind speeds are lower. The T* and q* plots also generally show a factor of two or better agreement among sites. T* and q* tend to be smaller in May, when the difference between the surface water temperature and the air temperature is at its minimum, and tend to be larger in December, when the difference between the surface water temperature and the air temperature is at its maximum (see Fig. 2 for the plot of monthly-averaged sea-air temperature differences). 4. Results of analysis of two case studies with COARE Time series of hourly averaged observations, COAREmodel calculations, and Eta-model simulations of boundary layer parameters at the observing sites in the Gulf of Mexico have been analyzed for several multi-day case study periods in different seasons. A few of the results from two case studies are discussed here. The first case study period that is presented in this paper is January 20e25, 2000, from buoy 42040, which is located several tens of kilometers to the southeast of the Mississippi River delta (see Fig. 1). Fig. 5 presents a time series of the hourlyaveraged air and water skin temperatures and wind speed for the time period. Because of the passage of two strong cold fronts during these five days, there are exceptionally large swings in wind and air temperature, with air temperatures
5e10 C cooler than the water skin temperatures for the first 2 days and for the last day and a half. Wind speeds are moderate to strong (about 5e15 m s1) during these periods. However, during the middle of the time period, when the wind shifted to the south, the air warms slowly to approach and even exceed the water temperature for over 12 h. The winds drop to nearly zero just before a cold front passage at about 03:00 h on 24 January, after which the air temperature drops 5 C in an hour and wind speed rapidly increases to 16 m s1. Fig. 6 shows the COARE-calculated and Eta-simulated sensible heat fluxes at the same buoy (42040) shown in Fig. 5 during this synoptically active time period. The COARE model is using the buoy-observed meteorological variables, and is seen to produce very large (for the ocean) sensible heat fluxes with magnitudes of about 200 W m2 during the beginning and ending periods, when the water-air temperature differences are very large (5e10 C) and the wind speeds are also large (about 10e15 m s1). However, during the 12e15-h period in the middle of the time series, when the air temperature exceeds the water temperature, the COARE-calculated sensible heat fluxes are negative (i.e., towards the water surface) with magnitudes of about 10 W m2. Eta model simulations are shown in Fig. 6 for forecast periods of 6e12 h and 30e36 h. It is seen that the Eta-based sensible heat fluxes are fairly close to (only about 30% larger than) the COARE calculated values during the periods with large air-water temperature differences. However, during the middle period, with warmer air temperatures, the Eta model simulates very slightly positive (upward) sensible heat fluxes (about 0e20 W m2), while the COARE-calculated sensible heat fluxes are slightly negative. This tendency is found for all sites and periods. That is, occasionally the buoys show periods with observed air temperatures warmer than water temperatures, leading to
546
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
COARE-calculated negative heat fluxes, while the Eta model is simulating positive (but small) heat fluxes. Evidently the Eta model’s parameterizations and assumptions tend to not produce significant time periods with negative heat fluxes. During the late spring, when the air temperature is observed to be greater than the water temperature about 20e40% of the time, and hence COARE is calculating negative heat fluxes, this can lead to long periods of mismatches in the signs of the COARE and Eta simulated sensible heat fluxes. Fortunately, during these time periods with the mismatches, the magnitudes of the sensible heat fluxes are fairly small and these differences have only minor effects on the surface heat budget. As discussed earlier, the latent heat flux usually overwhelms the sensible heat flux in the Gulf of Mexico. Although there is not space in this paper to present the complete set of results for the latent heat flux, there are relatively large values of about 600 W m2 during the post-cold front periods. These magnitudes of the latent heat flux are approximately equal to the solar energy flux at the surface at mid-day in the summer. The second case study period that is presented in this paper is September 18e20, 2001, for the South Marsh Island (SMI) oil platform, which is located about 150 km south of the Louisiana coast. Fig. 7 presents a time series of the hourly-averaged air and water skin temperatures and wind speed for that relatively warm period, which is included because it is a typical summertime synoptic situation with weak winds. Temperatures are fairly constant at about 30 C and the sea-air temperature difference persists at about 1e3 C during the period. Wind speed averages about 3 m s1 over the three days, but exceeds 5 m s1 for a few hours on September 18, and drops below 1 m s1 during the afternoon of September 19 and the late evening of September 20. For this time
period, the derived roughness length at the SMI site is about 0.00003 m most of the time, but increases, by a factor of about 20, to about 0.00055 m, during light wind periods when the wind speed drops below 1.5 m s1. This variation of roughness length is the opposite of what is expected in over-water boundary layers, and is thought to be caused by possible problems in user inputs and default assumptions for wave heights and periods during light winds in the revised COARE roughness algorithm. It is possible for relatively steep waves to result from the inputs, even though the wave heights are small. Fig. 8 shows the COARE-calculated and Eta-simulated hourly-averaged sensible heat fluxes and latent heat fluxes during the 18e20 September 2001 time period. In agreement with the general findings in the climatological study summarized in Section 3, the latent heat fluxes are about 5e10 times larger than the sensible heat fluxes, when averaged over the 3-day period. The COARE calculations and Eta simulations are seen to track each other fairly well (i.e., differences are less than a factor of two) most of the time. There are a few hours with large positive or negative differences between COARE and Eta, and these may be hours where the observed cloud cover (used in COARE) differs from the Eta-predicted cloud cover. The COARE-calculated and Eta-simulated hourly-averaged friction velocity, u*, and the stability parameter, z/L, were also compared for the 18e20 September time period. The friction velocities are relatively small, averaging about 0.2 m s1, consistent with the light winds and small roughness length. The variation in friction velocity closely tracks the variation in observed wind speed, since friction velocity is proportional to wind speed. About half of the time, the COARE u* calculations are less than the Eta simulations by about a factor of two. However, there is little bias the rest of the time, and there is better than factor of two agreement most of the time. The
Fig. 5. Observed air and water temperatures and observed wind speed at buoy 42040 (see Fig. 1 for location) for the case study period January 20 through 25, 2000.
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
547
Fig. 6. Estimated COARE and simulated Eta hourly sensible heat fluxes at buoy 42040 (see Fig. 1 for location) for the case study period January 20 through 25, 2000. Simulated Eta sensible heat fluxes were obtained from the 6- and 12-h and the 30- and 36-h forecasts and interpolated to hourly values between the 6- and 12h and the 30- and 36-h forecasts.
COARE-calculated z/L is typically about 5, reflecting the unstable environment expected for light wind conditions with a positive sea-air temperature gradient and with humid conditions. Note that L includes the influence of both the sensible and latent heat fluxes, since the major contributor to the overall buoyancy flux is the latent heat flux. 5. Comparison of winds from RWPs and from ETA/EDAS meteorological model simulations The locations of the six Radar Wind Profilers (RWPs) used during the study were shown in Fig. 1. The RWPs were LAP3000, 915-MHz boundary layer radars, with five deployed on oil/gas production platforms, and one deployed on the coast
(Fort Morgan). Two modes of data collection were employed at all RWP sites. The ‘‘low’’ mode represented 60-m resolution winds at heights from about 100 to 2000 m above platform level. The ‘‘high’’ mode represented 100-m resolution winds at heights from about 200 to about 4000 m above platform level. Because of the interference of the platform structures and the presence of sea clutter, few data were obtained at heights less than about 200 or 300 m. The Eta/EDAS meteorological model simulations of winds that were analyzed represent a grid-volume average obtained by bilinear interpolations of the 12-km Eta model output to a 40-km grid. Note that EDAS refers to the Eta Data Assimilation System. Vertical interpolation is performed on the 12-km Eta model (60 levels) to the EDAS grid (39 levels).
Fig. 7. Observations of hourly air and skin (water surface) temperatures and wind speed at the SMI platform (see Fig. 1 for location) for the case study period September 18 through 20, 2001.
548
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
Fig. 8. COARE and Eta latent and sensible heat fluxes at the SMI platform (see Fig. 1 for location) for the case study period September 18 through 20, 2001. Note: the four-digit number after Eta indicates the forecast period (i.e., 0612 is the 6- to 12-h forecast).
The RWP wind observations are nearly instantaneous values over a conical volume that is dependent on vertical resolution and beam width (approximately equal to height above ground). Therefore the volume represented by the RWP observation is much smaller than the volume represented by the Eta/EDAS simulation. The RWP data have been averaged over one hour for comparison with the Eta/EDAS simulations. Statistics on differences between RWP wind observations and Eta/EDAS wind simulations have been computed for those height bins containing at least 90% of the total number of RWP/Eta possible data pairs. Sample sizes at each height and each RWP location typically range from about 900 for the eastern sites associated with BAMP and operated for about 1 year, to about 2000 for the western sites, which have been in operation for over 3 years. Fig. 9 illustrates the results of the analysis, showing the wind speed and direction mean bias and mean standard deviation or SDV (i.e., the root-mean-square difference), for the entire data collection period at all sites. Biases are defined as the Eta/EDAS value minus the RWP value. The calculated standard deviations include the effects of the mean bias. In general, wind speeds are found to be overestimated by Eta/ EDAS by a slight amount, with more mean bias at the lower heights and with gradual improvement above 1000 m. The coastal land site, FTM, has the lowest wind speed mean bias, ranging from 0.5 to 0 m s1. BIP, WDP and VRM all have wind speed mean biases generally between 2 m s1 (near the surface) and 1 m s1 (at z ¼ 1000 m). The far offshore sites (SMI and DWP) have the highest wind speed mean biases of 3 m s1 (near the surface) to 2 m s1 (at z ¼ 1000 m), and 6 m s1 (near the surface) to 3 m s1 (at z ¼ 1000 m), respectively. These wind speed mean biases suggest that Eta/EDAS tends to overestimate boundary layer wind speeds in offshore areas of the Gulf of Mexico, with the difference increasing as distance from shore increases. The standard deviations of wind speed difference were 2e4 m s1 and indicate only slight decrease with altitude,
which is somewhat surprising as lower standard deviations are expected aloft because the upper-level flow patterns are known to vary less than flows near the surface (Seaman, 2000). Wind direction mean biases shown in Fig. 9 are between about 5 and þ10 at most sites and elevations, with higher wind direction biases observed (about 15e20 ) at heights below 1000 m at DWP. The mostly positive mean biases indicate that the Eta/EDAS wind directions are generally turned to the left with respect to the RWP wind directions. For example, with a 20 positive mean bias, if the RWP observed mean wind direction were 180 , then the Eta/EDAS-simulated mean wind direction would be 160 . The standard deviations of wind direction differences were generally about 25e45 and showed little variation with height. While the wind direction mean biases at offshore sites were similar to those at the near-shore sites, the standard deviations of the wind direction differences were generally about 40 at the offshore sites. These standard deviation values for wind speed and wind directions (24 m s1 and 20e50 ) found for the six RWPs in the Gulf of Mexico domain agree well with standard deviations reported by Seaman (2000) and Hanna and Yang (2001) for other models and other geographic domains. 6. Conclusions and recommendations The analysis is based on 3 years of observations of surface conditions and vertical profiles from several meteorological sites in the Gulf of Mexico (see Fig. 1). These data were used to investigate the over-water surface energy balance, and the climatology of latent heat fluxes, sensible heat fluxes, momentum fluxes, and moisture fluxes. Estimates of the scaling velocity (u*) and scaling temperature (T*) were studied. The widely used COARE algorithm was used for estimation of surface fluxes based on routine measurements. Threedimensional predicted fields of surface winds, heat and momentum fluxes, and wind profiles, from the National Center for Environmental Prediction’s (NCEP’s) Eta model, were
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
549
Fig. 9. Comparison of Radar Wind Profiler (RWP) observed wind speeds and directions with Eta/EDAS simulations for the total data collection period at all sites (see Fig. 1 for locations). Top: wind speed and direction mean bias. Bottom: mean standard deviation (STD). The bias is defined as positive if the Eta/EDAS value exceeds the RWP value. The standard deviation includes the effects of the mean bias.
compared with the observations from the RWPs and buoys. The analysis included (1) climatological studies of monthlyaveraged flux estimates from 11 sites, (2) case studies of two multi-day time periods, and (3) comparisons of RWPobserved and Eta-simulated wind speeds and directions. The climatological analysis of monthly averages showed that differences between the observed water ‘‘skin’’ and air temperatures were, on average, þ1 to þ3 C at most sites all year. The differences were less in late spring (about 0.5 C in May and June) and greater in late fall and early winter (about 2e5 C in December). The boundary layer was found to be unstable over 90% of the time. The fluxes and scaling parameters calculated by the COARE algorithm in the Gulf of Mexico are physically consistent with expectations and are similar in magnitude to the previous observations and COARE calculations for TOGA, which took place in the warm western Pacific Ocean near the equator. Calculated monthly average sensible heat fluxes in the Gulf of Mexico ranged from 10 to 30 W m2, typical of other over-water areas. Similarly, calculated monthly average latent heat fluxes ranged from 50 to 150 W m2, also typical of other over-water areas.
Both the latent and sensible heat fluxes were highest in December and lowest in May and June, although the yearly cycle in latent heat values is less pronounced. The calculated fluxes are generally in good agreement with the monthly average Eta model latent and sensible heat fluxes. The exception is the sensible heat flux in summer, where COARE calculates frequent negative values, whereas Eta simulates slightly positive values. The analysis focused on sensible heat fluxes, for which the 11 sites agree within about a factor of two in December, when the latent heat fluxes are at their largest, averaging 30e40 W m2. In the summer however, the sensible heat flux drops to less than 10 W m2, and is sometimes negative, causing poor agreement among the 11 sites. The COAREcalculated monthly average friction velocity, u*, for the buoys, the CMAN sites, and the oil platforms shows agreement among these sites well within a factor of two and often within 20%. This agreement is important because the monthly average friction velocity is the key scaling velocity for estimating transport speeds and dispersion rates. The monthly average Eta model friction velocity was usually within about 10e20% of the COARE-calculated friction velocity.
550
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
Two case studies were chosen for analysis of hourly variations over several days. The first case study, on January 20e25, 2000, was chosen because it was an extreme example of wide variations of synoptic conditions in the winter. Two strong cold fronts passed the area during the period, causing large sea-air temperature differences of 5e10 C and strong winds of 10e15 m s1. In between these two cold periods was a day when the air temperatures were warmer than the sea temperature and wind speeds were light. The COARE algorithm calculated hourly-averaged sensible heat fluxes of about 200 W m2 during the cold periods and about 10 W m2 during the warm periods. These were compared with sensible heat fluxes simulated by the Eta model, which were only about 30% larger than the COARE values during the cold periods, but were small but positive during the warm periods. The Eta model was found to produce much fewer negative sensible heat fluxes than the COARE calculations based on the observations. The second case study, on September 18e20, 2001, was chosen because it was a typical summertime period with light winds (averaging 3 m s1) and warm temperatures (averaging 30 C). As found elsewhere, the latent heat fluxes (averaging about 10 or 20 W m2) were consistently about five to ten times larger than the sensible heat fluxes (averaging about 50e100 W m2). The COARE-calculated and Eta-simulated sensible and latent heat fluxes tracked each other fairly well, with most within a factor of two. The COARE and Eta friction velocities, u*, also tracked each other well with about half of the hours showing agreement within a factor of two. The ratio of height to MonineObukhov length, L, was found to equal about 5 (unstable) most of the time. A slight problem was found with the COARE calculations of surface roughness length, zo, which was estimated to be about 20 times larger during light winds (less than about 1.5 m s1). The Eta/EDAS-simulated wind fields and the observed RWP winds from six sites were compared. Eta/EDAS is based on a combination of Eta model forecast winds and diagnostic interpolations of observed winds but does not include the RWP data. Focusing on the lowest 1000 or 2000 m of the boundary layer, the mean wind speed (WS) bias is less than 1 m s1 close to the shore but increases with offshore distance, so that the Eta/EDAS mean wind speed exceeds the RWP mean wind speed by 1e2 m s1 at 50 km offshore and by 2e6 m s1 at 100e200 km offshore. Mean wind direction (WD) bias is small, ranging from about 5 to 20 (note that, for a 20 WD bias, if the RWP WD were 180 , then the Eta/EDAS WD would be 160 ). Standard deviations of the differences (with mean bias removed) are 2e4 m s1 for WS and 20e50 for WD, in agreement with findings for other domains and models. However, these comparisons primarily focus on heights of a few hundred meters, since, due to sea clutter, the RWP seldom provided observations at heights less than about 200 m. The following recommendations are made for future studies:
Collocate radiometers and underwater temperature sensors to better determine the accuracy of the COARE-estimated warm-layer and cool-skin effect, and possibly devise corrections. Routinely operate a measurement system on an offshore platform that measures a range of meteorological parameters over the full depth of the boundary layer. Such a system could include a mini-Sodar, an RWP/RASS system, and a surface meteorological monitoring system on a platform to obtain wind and temperature measurements. The addition of the mini-Sodar would fill the measurement void that exists from 30 to 200 m, which is the layer that the RWPs are missing and where air quality concerns are maximized. Improve the COARE algorithms for shallow water conditions with light wind speeds, where surface roughness may be overpredicted, and slight diurnal variations are not accounted for. Develop a better understanding of the reasons for the bias between the Eta model and observed RWP winds at offshore locations, and use this knowledge to improve the Eta model parameterizations.
Acknowledgments This research was sponsored by the Minerals Management Service of the U.S. Department of the Interior. The authors appreciate the technical advice from Dr. Christopher Fairall, the developer of the COARE model, and the assistance given by numerous persons responsible for meteorological data collection on the oil platforms. References Black, T., 1994. The new NMC mesoscale Eta model: description and forecast examples. Weather Forecasting 9, 265e278. Brunke, M.A., Fairall, C.W., Zeng, X., 2003. Which bulk aerodynamic algorithms are least problematic in computing ocean surface turbulent fluxes? Journal of Climatology 16, 619e635. Fairall, C.W., Bradley, E.F., Rogers, D.P., Edson, J.B., Young, G.S., 1996. Bulk parameterization of air-sea fluxes for TOGA COARE. Journal of Geophysical Research 101, 3747e3764. Fairall, C.W., Bradley, E.F., Hare, J.E., Grachev, A.A., Edson, J.B., 2003. Bulk parameterization of air-sea fluxes: updates and verification for the COARE algorithm. Journal of Climatology 16, 571e591. Hanna, S.R., Yang, R., 2001. Evaluations of mesoscale model predictions of near-surface winds, temperature gradients, and mixing depths. Journal of Applied Meteorology 42, 453e466. Liu, W.T., Katsaros, K.B., Businger, J.A., 1979. Bulk parameterization of the air-sea exchange of heat and water vapor including the molecular constraints at the interface. Journal of Atmospheric Science 36, 1722e1735. Seaman, N.L., 2000. Meteorological modeling for air quality assessments. Atmospheric Environment 34, 2231e2259. Taylor, P.K., Yelland, M.J., 2000. The dependence of sea surface roughness on the height and steepness of the waves. Journal of Physical Oceanography 31, 572e590.
Analysis of three years of boundary layer observations over the Gulf of Mexico and its shores Steven R. Hanna a,*, Clinton P. MacDonald b, Mark Lilly b,1, Charles Knoderer b, C.H. Huang c a Hanna Consultants, 7 Crescent Avenue, Kennebunkport, ME 04046-7235, USA Sonoma Technology, Inc., 1360 Redwood Way, Suite C, Petaluma, CA 94954-1169, USA c Minerals Management Service, 1201 Elmwood Park Blvd., New Orleans, LA 70123, USA b
Received 1 June 2006; accepted 1 June 2006 Available online 6 September 2006
Abstract Boundary layer observations were made over the Gulf of Mexico over a 3-year period in order to develop and test methods for estimating surface fluxes and boundary layer wind fields. In addition to routinely available buoy and CMAN surface data, six 915 MHz radar wind profilers (RWPs) and RASS profilers were mounted on oil platforms and on the shore. Estimates of surface momentum, sensible heat, and latent heat fluxes have been made from the surface observations using the COARE software. Simulations by the National Weather Service’s Eta meteorological model are compared with the observations of surface fluxes and wind profiles. The boundary layer is found to be unstable over 90% of the time, and latent heat fluxes are about five to ten times larger than sensible heat fluxes, as usually found over tropical oceans. Eta model simulations of surface fluxes are within about 50% of COARE estimates of the fluxes based on surface observations. Most of the time, COAREderived fluxes at 11 sites are within a factor of two of each other at any given hour. In multi-day case studies, COARE calculations are found to agree with Eta model simulations of these fluxes and parameters within a factor of two most of the time. Eta model simulations of wind speeds in the boundary layer tend to exceed the RWP observations by 1e2 m s1 near shore and by 2e6 m s1 at distances of 100e200 km offshore. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: marine boundary layer; Gulf of Mexico meteorology; COARE; surface fluxes
1. Introduction This study was driven by the need to calculate the transport and dispersion of pollutants emitted from offshore oil and gas exploration and production activities in the Gulf of Mexico. Much uncertainty exists concerning the atmospheric boundary layer in the region, due to a lack of detailed observations. Complications arise due to space and time variations caused by variable underlying water temperatures and the effects of mesoscale atmospheric phenomena. The current study has involved instrumentation of six offshore oil/gas production platforms to obtain boundary layer observations, including * Corresponding author. E-mail address: [email protected] (S.R. Hanna). 1 Present address: Northrup Grumman Information Technology, 6940 Kings Highway, Alexandria, VA 22319, USA. 0272-7714/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.ecss.2006.06.005
915-MHz Radar Wind Profilers (RWPs), 2-kHz Radio Acoustic Sounding Systems (RASS), and near-surface routine meteorology instruments. Fig. 1 shows a map of the Gulf of Mexico region and indicates the locations of the various observing sites. Two of the profiler sites have operated from May 1998 through September 2001 and four of the sites, sponsored by the Breton Aerometric Monitoring Program (BAMP), operated from October 2000 through September 2001. The RWP measures winds and the RASS measures virtual temperatures between heights of about 100 m and a few kilometers. The near-surface observations at the oil platforms include sea surface temperature as well as wind speed, wind direction, air temperature, and mixing ratio at a reference elevation, zr, of about 25 m (i.e., the platform elevation). These new data, in addition to the traditional data collected by buoys and by the Coastal Marine Automated Network (CMAN), which are available from the National Climatic Data Center (NCDC),
542
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
Fig. 1. Map of the study region in the western and central Gulf of Mexico showing locations of CMAN sites, buoys, and profilers. The two ‘‘MMS’’ profilers operated for about 3 years, while the four BAMP profilers operated for about 1 year. The numbers in parentheses are the water depth in meters.
have been analyzed to investigate the over-water surface energy balance and boundary layer structure for both steady-state horizontally homogeneous conditions and for conditions variable in time and space. These three-dimensional, timedependent fields are being used for analysis of transport and dispersion from over-water sources. The Tropical Ocean Global Atmosphere-Coupled Oceanic Atmospheric Response Experiment (TOGA-COARE) marine boundary layer algorithms (Fairall et al., 1996, 2003) have been used to calculate hourly-averaged and monthly-averaged fundamental boundary layer scaling parameters such as the surface roughness length (zo), the friction velocity (u*), and the MonineObukhov length (L), in addition to the latent heat flux (Qh) and the sensible heat flux (Qs). From these parameters, the mixing depth (h), and the vertical profiles of wind speed (u), temperature (T ), and water vapor mixing ratio (q) can be estimated for input to transport and dispersion models. The basic structure of the COARE marine boundary-layer model is an outgrowth of the LiueKatsaroseBusinger (LKB) (Liu et al., 1979) method. Version 3.0 of COARE, used in the current analysis and described by Fairall et al. (2003), incorporates wave height and period data, in order to increase the accuracy of the estimates of surface fluxes and scaling parameters over shallow areas (Taylor and Yelland, 2000). This allows COARE to account for differences in the characteristics of waves from the deep ocean to the shallow coastal waters. COARE’s accuracy was shown in the model evaluation exercise carried out by Brunke et al. (2003), who evaluated 12 bulk aerodynamic algorithms for computing ocean surface turbulent fluxes with observations from 12 ship cruises over the tropical and midlatitude oceans. They found that COARE
was one of the group of four algorithms that showed the best performance. In the sections below, the outputs of the COARE program (e.g., boundary layer scaling parameters and energy fluxes) are compared to observations and to simulations by the National Center for Environmental Prediction (NCEP) Eta and Eta Data Assimilation System (EDAS) numerical weather prediction modeling system (Black, 1994). In addition, some comparisons are presented of the observed RWP winds and the Eta/EDAS modeled winds. This study focuses on the Western and Central Gulf of Mexico, west of the FloridaeAlabama border. Additional MMS studies (not discussed here) are focusing on the eastern Gulf of Mexico region. 2. Application of the COARE model The Coupled Oceanic Atmospheric Response Experiment (COARE) model (Fairall et al., 1996, 2003) allows estimation of over water boundary layer parameters such as surface fluxes of momentum (t), sensible heat (Qs), and latent heat (Qh), and requires the following input data: Site location and time Wind speed (u), air temperature (Ta), and relative humidity (RH) within the atmospheric surface layer at a reference height zr Sea surface skin temperature (Ts) (or water temperature near the sea surface plus an estimate of solar and net longwave radiation) Mixing height (h)
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
If the water temperature at some level near (i.e., a short distance below) the sea surface is used, then solar and downwelling longwave radiation fluxes need to be estimated from some alternate source in order to correct this temperature to a skin (sea surface) temperature. Precipitation data are not required, but if available, can be used by COARE to estimate the precipitation contribution to the kinetic and thermal energy balance equations. Wave height and period data are not required, but, when available, can be used with Version 3.0 of COARE (Fairall et al., 2003) to account for the different wave structures and theoretically improve the accuracy of the estimates of surface fluxes and scaling parameters over shallow ocean areas. The COARE equations and their justifications are not discussed here, since the focus of the current study is the application of COARE to the extensive data sets collected in the Gulf of Mexico. Fairall et al. (1996, 2003) discuss the details of the COARE equations, and Brunke et al. (2003) show that COARE is one of the better-performing group of models when compared with observations from 12 ship cruises in tropical and midlatitude oceans. Data were acquired and processed from the following sites: 1998 through 2001 offshore buoy data from seven sites, shoreline CMAN station data from five sites, and data collected as part of this project on the Vermillion (VRM), South Marsh Island (SMI), Breton Island (BI), Fort Morgan (FM), Deep Water Platform (DWP), and Shallow Water Platform (SWP) oil/gas production platforms. Fig. 1 shows the locations of the offshore buoys, the CMAN stations, the oil/gas production platforms and the Fort Morgan site (with surface observations plus profilers). The CMAN stations are located on or adjacent to very shallow water on the coast. Since the COARE model is not currently designed to use data collected in such areas, the data from the CMAN sites were used only in test exercises to evaluate how COARE would respond compared to open water sites. The data collected at the platforms generally meet the COARE input data requirements, although the elevations (20e25 m) are relatively high compared to most of the elevations used in development of the model. The data collected at buoys meet most of the COARE data requirements, with the exception that solar and longwave radiation were not observed. Instead, for the buoys, radiation fluxes were estimated using 6-hourly Eta model cloud forecasts and sun elevation data, from which COARE calculates the water skin temperatures from water temperatures at depths of about 0.5e1.0 m. The climatological part of the study is discussed first. Using the hourly meteorological data collected for over 3 years at 11 sites where the data passed QA procedures, COARE was used to calculate hourly sensible heat flux (Qs), latent heat flux (Qh), surface stress (t), friction velocity (u*), temperature scaling parameter (T*), relative humidity scaling parameter (q*), zr/L, and roughness length (zo). Note that zr is the reference height for the measurement and L is the MonineObukhov length. Monthly statistics were then calculated from these hourly values. Monthly averages were not computed if more than 80% of the data in a given month were missing.
543
The case study part of the analysis is discussed second. In this analysis, time series of hourly averages of selected meteorological and derived boundary layer parameters were calculated using COARE, for two multi-day periods during different seasons. The analysis includes comparisons of derived outputs from COARE with simulations of the Eta model for sensible and latent heat flux, and friction velocity. The final part of the analysis involves comparisons of the winds observed by the RWPs at elevations of a few hundred meters with simulations of the Eta model. Mean biases and mean-square errors are presented for the total 3-year period. 3. Results of climatological study with COARE The key results of the Gulf of Mexico boundary layer climatological study (comparisons of monthly averages at 11 sites) with COARE are briefly summarized in this section. The fluxes and scalar parameters calculated by the COARE algorithm for the 3-year observing period in the Gulf of Mexico are physically consistent with our intuitive expectations, and are similar to observations and COARE calculations for the TOGA field study, which took place in the warm Western Pacific Ocean near the equator (Fairall et al., 1996). As expected, the COARE-calculated fluxes in the Gulf of Mexico tend to be strongly related to the sea-air temperature difference. Fig. 2 contains monthly-averaged observed sea-air temperature difference over the 3-year period at seven buoys, two CMAN sites, and the VRM and SMI oil platforms (see Fig. 1 for the locations). An annual variation is seen with maximum differences (about 2e5 C) in December, and minimum differences (about 0.5 C) in May and June. Our analysis shows that the sea-air temperature difference is positive over 90% of the time, implying that well-mixed conditions occur most of the time. The COARE-calculated monthly-average sensible heat fluxes, Qs, over the 3-year period are plotted in Fig. 3 for the same 11 sites used in Fig. 2. Calculated annual-averaged sensible heat fluxes in the Gulf of Mexico average about 10e30 W m2, typical of other over-water sites (Fairall et al., 1996, 2003). Slightly higher sensible heat fluxes (averaging about 30e40 W m2) are calculated in the winter months, especially in December and January, due to larger sea-air temperature differences and stronger winds during the winter. The sensible heat fluxes at the 11 sites tend to agree within a factor of two most of the time during the period from September through March. Larger differences between the 11 sites are seen in Fig. 3 during the summer months, when there are smaller sea-air temperature differences and hence the sensible heat fluxes are relatively small (about 10 W m2 or less) and variable, and are sometimes very near zero or are negative. The COARE-calculated monthly average latent heat fluxes, Qh, are shown in Fig. 4 for the same 11 sites used in Fig. 3. The monthly-averaged latent heat fluxes in Fig. 4 range from about 30 to 260 W m2. The annual average latent heat flux over all 11 sites averages about 50e150 W m2, again typical of other over-water sites. Calculated latent heat fluxes are therefore, on average, about five times larger than the
544
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
Fig. 2. Monthly averaged observed skin (water surface) temperature minus air temperature, for 11 sites in the Gulf of Mexico (see Fig. 1 for locations) for May 1998 through October 2001. The skin temperature was observed only at the platforms (SMI and VRM) and was estimated by COARE from the observed water temperature (at a depth just below the surface) at the other sites.
calculated sensible heat fluxes. This approximate ratio is found at other open-water sites at this latitude (e.g., Fairall et al., 1996). The magnitudes of the fluxes are generally greater at all 11 sites in the late fall months, likely due to the higher water temperatures and the larger sea-air temperature differences and wind speeds at that time. As is well known, the annual
maximum and minimum water temperatures have a time lag of a few months after the maximum and minimum air temperatures. Latent heat fluxes over the Gulf of Mexico are seldom negative due to the fact that the air at the surface is obviously always saturated with water vapor and the sea-air temperature difference is positive 90% of the time.
Fig. 3. Monthly averages of COARE-estimated sensible heat fluxes, for 11 sites in the Gulf of Mexico (see Fig. 1 for locations) for May 1998 through October 2001.
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
545
Fig. 4. Monthly averages of COARE-estimated latent heat fluxes, for 11 sites in the Gulf of Mexico (see Fig. 1 for locations) for May 1998 through October 2001.
Plots of the COARE-calculated monthly average friction velocity, u*, temperature scaling parameter, T*, and humidity scaling parameter, q*, were made but are not given in the current paper due to space limitations. The plots of u* show an agreement among the sites well within a factor of two and often within 20%. This agreement is important because u* is the key scaling velocity for estimating transport speeds and dispersion rates. u* is calculated to be slightly lower from May through August when wind speeds are lower. The T* and q* plots also generally show a factor of two or better agreement among sites. T* and q* tend to be smaller in May, when the difference between the surface water temperature and the air temperature is at its minimum, and tend to be larger in December, when the difference between the surface water temperature and the air temperature is at its maximum (see Fig. 2 for the plot of monthly-averaged sea-air temperature differences). 4. Results of analysis of two case studies with COARE Time series of hourly averaged observations, COAREmodel calculations, and Eta-model simulations of boundary layer parameters at the observing sites in the Gulf of Mexico have been analyzed for several multi-day case study periods in different seasons. A few of the results from two case studies are discussed here. The first case study period that is presented in this paper is January 20e25, 2000, from buoy 42040, which is located several tens of kilometers to the southeast of the Mississippi River delta (see Fig. 1). Fig. 5 presents a time series of the hourlyaveraged air and water skin temperatures and wind speed for the time period. Because of the passage of two strong cold fronts during these five days, there are exceptionally large swings in wind and air temperature, with air temperatures
5e10 C cooler than the water skin temperatures for the first 2 days and for the last day and a half. Wind speeds are moderate to strong (about 5e15 m s1) during these periods. However, during the middle of the time period, when the wind shifted to the south, the air warms slowly to approach and even exceed the water temperature for over 12 h. The winds drop to nearly zero just before a cold front passage at about 03:00 h on 24 January, after which the air temperature drops 5 C in an hour and wind speed rapidly increases to 16 m s1. Fig. 6 shows the COARE-calculated and Eta-simulated sensible heat fluxes at the same buoy (42040) shown in Fig. 5 during this synoptically active time period. The COARE model is using the buoy-observed meteorological variables, and is seen to produce very large (for the ocean) sensible heat fluxes with magnitudes of about 200 W m2 during the beginning and ending periods, when the water-air temperature differences are very large (5e10 C) and the wind speeds are also large (about 10e15 m s1). However, during the 12e15-h period in the middle of the time series, when the air temperature exceeds the water temperature, the COARE-calculated sensible heat fluxes are negative (i.e., towards the water surface) with magnitudes of about 10 W m2. Eta model simulations are shown in Fig. 6 for forecast periods of 6e12 h and 30e36 h. It is seen that the Eta-based sensible heat fluxes are fairly close to (only about 30% larger than) the COARE calculated values during the periods with large air-water temperature differences. However, during the middle period, with warmer air temperatures, the Eta model simulates very slightly positive (upward) sensible heat fluxes (about 0e20 W m2), while the COARE-calculated sensible heat fluxes are slightly negative. This tendency is found for all sites and periods. That is, occasionally the buoys show periods with observed air temperatures warmer than water temperatures, leading to
546
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
COARE-calculated negative heat fluxes, while the Eta model is simulating positive (but small) heat fluxes. Evidently the Eta model’s parameterizations and assumptions tend to not produce significant time periods with negative heat fluxes. During the late spring, when the air temperature is observed to be greater than the water temperature about 20e40% of the time, and hence COARE is calculating negative heat fluxes, this can lead to long periods of mismatches in the signs of the COARE and Eta simulated sensible heat fluxes. Fortunately, during these time periods with the mismatches, the magnitudes of the sensible heat fluxes are fairly small and these differences have only minor effects on the surface heat budget. As discussed earlier, the latent heat flux usually overwhelms the sensible heat flux in the Gulf of Mexico. Although there is not space in this paper to present the complete set of results for the latent heat flux, there are relatively large values of about 600 W m2 during the post-cold front periods. These magnitudes of the latent heat flux are approximately equal to the solar energy flux at the surface at mid-day in the summer. The second case study period that is presented in this paper is September 18e20, 2001, for the South Marsh Island (SMI) oil platform, which is located about 150 km south of the Louisiana coast. Fig. 7 presents a time series of the hourly-averaged air and water skin temperatures and wind speed for that relatively warm period, which is included because it is a typical summertime synoptic situation with weak winds. Temperatures are fairly constant at about 30 C and the sea-air temperature difference persists at about 1e3 C during the period. Wind speed averages about 3 m s1 over the three days, but exceeds 5 m s1 for a few hours on September 18, and drops below 1 m s1 during the afternoon of September 19 and the late evening of September 20. For this time
period, the derived roughness length at the SMI site is about 0.00003 m most of the time, but increases, by a factor of about 20, to about 0.00055 m, during light wind periods when the wind speed drops below 1.5 m s1. This variation of roughness length is the opposite of what is expected in over-water boundary layers, and is thought to be caused by possible problems in user inputs and default assumptions for wave heights and periods during light winds in the revised COARE roughness algorithm. It is possible for relatively steep waves to result from the inputs, even though the wave heights are small. Fig. 8 shows the COARE-calculated and Eta-simulated hourly-averaged sensible heat fluxes and latent heat fluxes during the 18e20 September 2001 time period. In agreement with the general findings in the climatological study summarized in Section 3, the latent heat fluxes are about 5e10 times larger than the sensible heat fluxes, when averaged over the 3-day period. The COARE calculations and Eta simulations are seen to track each other fairly well (i.e., differences are less than a factor of two) most of the time. There are a few hours with large positive or negative differences between COARE and Eta, and these may be hours where the observed cloud cover (used in COARE) differs from the Eta-predicted cloud cover. The COARE-calculated and Eta-simulated hourly-averaged friction velocity, u*, and the stability parameter, z/L, were also compared for the 18e20 September time period. The friction velocities are relatively small, averaging about 0.2 m s1, consistent with the light winds and small roughness length. The variation in friction velocity closely tracks the variation in observed wind speed, since friction velocity is proportional to wind speed. About half of the time, the COARE u* calculations are less than the Eta simulations by about a factor of two. However, there is little bias the rest of the time, and there is better than factor of two agreement most of the time. The
Fig. 5. Observed air and water temperatures and observed wind speed at buoy 42040 (see Fig. 1 for location) for the case study period January 20 through 25, 2000.
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
547
Fig. 6. Estimated COARE and simulated Eta hourly sensible heat fluxes at buoy 42040 (see Fig. 1 for location) for the case study period January 20 through 25, 2000. Simulated Eta sensible heat fluxes were obtained from the 6- and 12-h and the 30- and 36-h forecasts and interpolated to hourly values between the 6- and 12h and the 30- and 36-h forecasts.
COARE-calculated z/L is typically about 5, reflecting the unstable environment expected for light wind conditions with a positive sea-air temperature gradient and with humid conditions. Note that L includes the influence of both the sensible and latent heat fluxes, since the major contributor to the overall buoyancy flux is the latent heat flux. 5. Comparison of winds from RWPs and from ETA/EDAS meteorological model simulations The locations of the six Radar Wind Profilers (RWPs) used during the study were shown in Fig. 1. The RWPs were LAP3000, 915-MHz boundary layer radars, with five deployed on oil/gas production platforms, and one deployed on the coast
(Fort Morgan). Two modes of data collection were employed at all RWP sites. The ‘‘low’’ mode represented 60-m resolution winds at heights from about 100 to 2000 m above platform level. The ‘‘high’’ mode represented 100-m resolution winds at heights from about 200 to about 4000 m above platform level. Because of the interference of the platform structures and the presence of sea clutter, few data were obtained at heights less than about 200 or 300 m. The Eta/EDAS meteorological model simulations of winds that were analyzed represent a grid-volume average obtained by bilinear interpolations of the 12-km Eta model output to a 40-km grid. Note that EDAS refers to the Eta Data Assimilation System. Vertical interpolation is performed on the 12-km Eta model (60 levels) to the EDAS grid (39 levels).
Fig. 7. Observations of hourly air and skin (water surface) temperatures and wind speed at the SMI platform (see Fig. 1 for location) for the case study period September 18 through 20, 2001.
548
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
Fig. 8. COARE and Eta latent and sensible heat fluxes at the SMI platform (see Fig. 1 for location) for the case study period September 18 through 20, 2001. Note: the four-digit number after Eta indicates the forecast period (i.e., 0612 is the 6- to 12-h forecast).
The RWP wind observations are nearly instantaneous values over a conical volume that is dependent on vertical resolution and beam width (approximately equal to height above ground). Therefore the volume represented by the RWP observation is much smaller than the volume represented by the Eta/EDAS simulation. The RWP data have been averaged over one hour for comparison with the Eta/EDAS simulations. Statistics on differences between RWP wind observations and Eta/EDAS wind simulations have been computed for those height bins containing at least 90% of the total number of RWP/Eta possible data pairs. Sample sizes at each height and each RWP location typically range from about 900 for the eastern sites associated with BAMP and operated for about 1 year, to about 2000 for the western sites, which have been in operation for over 3 years. Fig. 9 illustrates the results of the analysis, showing the wind speed and direction mean bias and mean standard deviation or SDV (i.e., the root-mean-square difference), for the entire data collection period at all sites. Biases are defined as the Eta/EDAS value minus the RWP value. The calculated standard deviations include the effects of the mean bias. In general, wind speeds are found to be overestimated by Eta/ EDAS by a slight amount, with more mean bias at the lower heights and with gradual improvement above 1000 m. The coastal land site, FTM, has the lowest wind speed mean bias, ranging from 0.5 to 0 m s1. BIP, WDP and VRM all have wind speed mean biases generally between 2 m s1 (near the surface) and 1 m s1 (at z ¼ 1000 m). The far offshore sites (SMI and DWP) have the highest wind speed mean biases of 3 m s1 (near the surface) to 2 m s1 (at z ¼ 1000 m), and 6 m s1 (near the surface) to 3 m s1 (at z ¼ 1000 m), respectively. These wind speed mean biases suggest that Eta/EDAS tends to overestimate boundary layer wind speeds in offshore areas of the Gulf of Mexico, with the difference increasing as distance from shore increases. The standard deviations of wind speed difference were 2e4 m s1 and indicate only slight decrease with altitude,
which is somewhat surprising as lower standard deviations are expected aloft because the upper-level flow patterns are known to vary less than flows near the surface (Seaman, 2000). Wind direction mean biases shown in Fig. 9 are between about 5 and þ10 at most sites and elevations, with higher wind direction biases observed (about 15e20 ) at heights below 1000 m at DWP. The mostly positive mean biases indicate that the Eta/EDAS wind directions are generally turned to the left with respect to the RWP wind directions. For example, with a 20 positive mean bias, if the RWP observed mean wind direction were 180 , then the Eta/EDAS-simulated mean wind direction would be 160 . The standard deviations of wind direction differences were generally about 25e45 and showed little variation with height. While the wind direction mean biases at offshore sites were similar to those at the near-shore sites, the standard deviations of the wind direction differences were generally about 40 at the offshore sites. These standard deviation values for wind speed and wind directions (24 m s1 and 20e50 ) found for the six RWPs in the Gulf of Mexico domain agree well with standard deviations reported by Seaman (2000) and Hanna and Yang (2001) for other models and other geographic domains. 6. Conclusions and recommendations The analysis is based on 3 years of observations of surface conditions and vertical profiles from several meteorological sites in the Gulf of Mexico (see Fig. 1). These data were used to investigate the over-water surface energy balance, and the climatology of latent heat fluxes, sensible heat fluxes, momentum fluxes, and moisture fluxes. Estimates of the scaling velocity (u*) and scaling temperature (T*) were studied. The widely used COARE algorithm was used for estimation of surface fluxes based on routine measurements. Threedimensional predicted fields of surface winds, heat and momentum fluxes, and wind profiles, from the National Center for Environmental Prediction’s (NCEP’s) Eta model, were
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
549
Fig. 9. Comparison of Radar Wind Profiler (RWP) observed wind speeds and directions with Eta/EDAS simulations for the total data collection period at all sites (see Fig. 1 for locations). Top: wind speed and direction mean bias. Bottom: mean standard deviation (STD). The bias is defined as positive if the Eta/EDAS value exceeds the RWP value. The standard deviation includes the effects of the mean bias.
compared with the observations from the RWPs and buoys. The analysis included (1) climatological studies of monthlyaveraged flux estimates from 11 sites, (2) case studies of two multi-day time periods, and (3) comparisons of RWPobserved and Eta-simulated wind speeds and directions. The climatological analysis of monthly averages showed that differences between the observed water ‘‘skin’’ and air temperatures were, on average, þ1 to þ3 C at most sites all year. The differences were less in late spring (about 0.5 C in May and June) and greater in late fall and early winter (about 2e5 C in December). The boundary layer was found to be unstable over 90% of the time. The fluxes and scaling parameters calculated by the COARE algorithm in the Gulf of Mexico are physically consistent with expectations and are similar in magnitude to the previous observations and COARE calculations for TOGA, which took place in the warm western Pacific Ocean near the equator. Calculated monthly average sensible heat fluxes in the Gulf of Mexico ranged from 10 to 30 W m2, typical of other over-water areas. Similarly, calculated monthly average latent heat fluxes ranged from 50 to 150 W m2, also typical of other over-water areas.
Both the latent and sensible heat fluxes were highest in December and lowest in May and June, although the yearly cycle in latent heat values is less pronounced. The calculated fluxes are generally in good agreement with the monthly average Eta model latent and sensible heat fluxes. The exception is the sensible heat flux in summer, where COARE calculates frequent negative values, whereas Eta simulates slightly positive values. The analysis focused on sensible heat fluxes, for which the 11 sites agree within about a factor of two in December, when the latent heat fluxes are at their largest, averaging 30e40 W m2. In the summer however, the sensible heat flux drops to less than 10 W m2, and is sometimes negative, causing poor agreement among the 11 sites. The COAREcalculated monthly average friction velocity, u*, for the buoys, the CMAN sites, and the oil platforms shows agreement among these sites well within a factor of two and often within 20%. This agreement is important because the monthly average friction velocity is the key scaling velocity for estimating transport speeds and dispersion rates. The monthly average Eta model friction velocity was usually within about 10e20% of the COARE-calculated friction velocity.
550
S.R. Hanna et al. / Estuarine, Coastal and Shelf Science 70 (2006) 541e550
Two case studies were chosen for analysis of hourly variations over several days. The first case study, on January 20e25, 2000, was chosen because it was an extreme example of wide variations of synoptic conditions in the winter. Two strong cold fronts passed the area during the period, causing large sea-air temperature differences of 5e10 C and strong winds of 10e15 m s1. In between these two cold periods was a day when the air temperatures were warmer than the sea temperature and wind speeds were light. The COARE algorithm calculated hourly-averaged sensible heat fluxes of about 200 W m2 during the cold periods and about 10 W m2 during the warm periods. These were compared with sensible heat fluxes simulated by the Eta model, which were only about 30% larger than the COARE values during the cold periods, but were small but positive during the warm periods. The Eta model was found to produce much fewer negative sensible heat fluxes than the COARE calculations based on the observations. The second case study, on September 18e20, 2001, was chosen because it was a typical summertime period with light winds (averaging 3 m s1) and warm temperatures (averaging 30 C). As found elsewhere, the latent heat fluxes (averaging about 10 or 20 W m2) were consistently about five to ten times larger than the sensible heat fluxes (averaging about 50e100 W m2). The COARE-calculated and Eta-simulated sensible and latent heat fluxes tracked each other fairly well, with most within a factor of two. The COARE and Eta friction velocities, u*, also tracked each other well with about half of the hours showing agreement within a factor of two. The ratio of height to MonineObukhov length, L, was found to equal about 5 (unstable) most of the time. A slight problem was found with the COARE calculations of surface roughness length, zo, which was estimated to be about 20 times larger during light winds (less than about 1.5 m s1). The Eta/EDAS-simulated wind fields and the observed RWP winds from six sites were compared. Eta/EDAS is based on a combination of Eta model forecast winds and diagnostic interpolations of observed winds but does not include the RWP data. Focusing on the lowest 1000 or 2000 m of the boundary layer, the mean wind speed (WS) bias is less than 1 m s1 close to the shore but increases with offshore distance, so that the Eta/EDAS mean wind speed exceeds the RWP mean wind speed by 1e2 m s1 at 50 km offshore and by 2e6 m s1 at 100e200 km offshore. Mean wind direction (WD) bias is small, ranging from about 5 to 20 (note that, for a 20 WD bias, if the RWP WD were 180 , then the Eta/EDAS WD would be 160 ). Standard deviations of the differences (with mean bias removed) are 2e4 m s1 for WS and 20e50 for WD, in agreement with findings for other domains and models. However, these comparisons primarily focus on heights of a few hundred meters, since, due to sea clutter, the RWP seldom provided observations at heights less than about 200 m. The following recommendations are made for future studies:
Collocate radiometers and underwater temperature sensors to better determine the accuracy of the COARE-estimated warm-layer and cool-skin effect, and possibly devise corrections. Routinely operate a measurement system on an offshore platform that measures a range of meteorological parameters over the full depth of the boundary layer. Such a system could include a mini-Sodar, an RWP/RASS system, and a surface meteorological monitoring system on a platform to obtain wind and temperature measurements. The addition of the mini-Sodar would fill the measurement void that exists from 30 to 200 m, which is the layer that the RWPs are missing and where air quality concerns are maximized. Improve the COARE algorithms for shallow water conditions with light wind speeds, where surface roughness may be overpredicted, and slight diurnal variations are not accounted for. Develop a better understanding of the reasons for the bias between the Eta model and observed RWP winds at offshore locations, and use this knowledge to improve the Eta model parameterizations.
Acknowledgments This research was sponsored by the Minerals Management Service of the U.S. Department of the Interior. The authors appreciate the technical advice from Dr. Christopher Fairall, the developer of the COARE model, and the assistance given by numerous persons responsible for meteorological data collection on the oil platforms. References Black, T., 1994. The new NMC mesoscale Eta model: description and forecast examples. Weather Forecasting 9, 265e278. Brunke, M.A., Fairall, C.W., Zeng, X., 2003. Which bulk aerodynamic algorithms are least problematic in computing ocean surface turbulent fluxes? Journal of Climatology 16, 619e635. Fairall, C.W., Bradley, E.F., Rogers, D.P., Edson, J.B., Young, G.S., 1996. Bulk parameterization of air-sea fluxes for TOGA COARE. Journal of Geophysical Research 101, 3747e3764. Fairall, C.W., Bradley, E.F., Hare, J.E., Grachev, A.A., Edson, J.B., 2003. Bulk parameterization of air-sea fluxes: updates and verification for the COARE algorithm. Journal of Climatology 16, 571e591. Hanna, S.R., Yang, R., 2001. Evaluations of mesoscale model predictions of near-surface winds, temperature gradients, and mixing depths. Journal of Applied Meteorology 42, 453e466. Liu, W.T., Katsaros, K.B., Businger, J.A., 1979. Bulk parameterization of the air-sea exchange of heat and water vapor including the molecular constraints at the interface. Journal of Atmospheric Science 36, 1722e1735. Seaman, N.L., 2000. Meteorological modeling for air quality assessments. Atmospheric Environment 34, 2231e2259. Taylor, P.K., Yelland, M.J., 2000. The dependence of sea surface roughness on the height and steepness of the waves. Journal of Physical Oceanography 31, 572e590.